As semiconductor devices are becoming smaller and on-chip device density is correspondingly increasing, both signal delays due to capacitive coupling and crosstalk between closely spaced metal lines are increasing. These problems are exacerbated by the need to keep conductor lines as short as possible in order to minimize transmission delays, thus requiring multilevel wiring schemes for the chip. The problems have been ameliorated to some extent by the switch to copper metallurgy, but as feature sizes go below 0.25 μm, this alone will not provide a solution. The use of an insulator with a lower dielectric constant than the currently used SiO2 (k=3.9–4.2) would also, clearly, improve the situation. Current integration demands for insulators used with, for example, Al(Cu) wiring, also require thermal stabilities in excess of 400° C., good mechanical properties, resistance to crack generation and propagation, low defect densities, low water uptake, chemical resistance, processability by photolithographic techniques and gas phase etching procedures, and capacity for planarization.
Accordingly, considerable attention has focused on the replacement of silicon dioxide with new materials, particularly materials having lower dielectric constants, since both capacitive delays and power consumption depend on the dielectric constant of the insulator. This is not a simple matter given the complexities and demands of current semiconductor integration processes. Of the existing materials with demonstrated ultra-low dielectric constants, the highly fluorinated materials (e.g., Teflon) have the longest history. For example, attempts have been made to reduce the dielectric constant of polyimides by incorporating perfluoroalkyl-containing comonomers into the polymer structure (see, e.g., Haidar et al. (1991) Mater. Res. Soc. Symp. Proc. 227:35; Critchlen et al. (1972) J. Polym. Sci. A-1 10:1789; and Harris et al. (1991) Mater. Res. Soc. Symp. Proc. 227:3). The synthesis of polyimides based on 9,9-disubstituted xanthene dianhydrides, e.g., 6FXDA/6FDA (9,9-bis(trifluoromethoxy)xanthenetetracarboxylic dianhydride/2,2-bis(4-aminophenyl)-1,1,1,3,3,3-hexafluoropropane), as well as polyimides based on the TFMOB monomer (2,2′-bis(trifluoromethyl)benzidine), has been reported (see Muraka (March 1996) Solid State Tech 83 and Jang et al. (1994) Mater. Res. Soc. Symp. Proc. 337:25). Although fluorinated alkane polymers such as these have the lowest dielectric constants of any homogeneous polymers and substitution with fluorine decreases the dielectric constant of any polymer, there are many liabilities. Current integration requirements call for exceptional thermal stability at temperatures in the range of 400–450° C. This temperature region is a problem for most organic polymers, and particularly for many fluorocarbons. Also, adhesion of fluorinated materials (self-adhesion, adhesion to metals, dielectrics, ceramics, etc.) is a problem without some prior surface pretreatment. Further, the stability of fluorinated materials due to interaction with interconnect metallurgy at elevated temperatures is problematic. The mechanical properties of many known fluorinated materials (particularly fluoroalkanes) are not ideal; they usually have large thermal expansion coefficients and are intrinsically soft materials. The latter creates a problem for chemical mechanical polishing (CMP) procedures. Finally, the methodology to develop other highly fluorinated materials such as fluorinated polyimides is limited by synthetic difficulties associated with the incorporation of a substantial number of pendant perfluoroalkyl groups.
Attempts have been made to reduce the dielectric constant of such materials through the introduction of kinks and conjugation-interrupting linkages in the polymer backbone to lower molecular polarizability and reduce chain-chain interactions (St. Clair et al. (1988) Proc. Amer. Chem. Soc. Div. Polym. Mater. Sci. Eng. 59:28). A more viable approach, however, has been controlled introduction of porosity into existing low dielectric constant materials.
Generation of porous polymer foams substantially reduces the dielectric constant of the material while maintaining the desired thermal and mechanical properties of the base (or “host”) polymer. The reduction in dielectric constant is achieved by incorporating air voids, as air has a dielectric constant of 1.01. The advantage of a foam approach is illustrated in Hedrick et al. (1995) Polymer 36:2685, which illustrates in graph form a Maxwell-Garnett model of composite structures based on a matrix polymer having an initial dielectric constant of 2.8. Incorporation of a second phase of dielectric constant 1.01, as with the introduction of air-filled pores in a foam, causes a dramatic reduction in the dielectric constant. However, foams provide a unique set of problems for dielectric applications. The pore size must be much smaller than both the film thickness and any microelectronic device features. In addition, the pores are ideally closed cell, i.e., the connectivity between the pores must be minimal to prevent the diffusion of reactive contaminants. Finally, the volume fraction of the voids must be as high as possible to achieve the lowest possible dielectric constant. All of these features can alter the mechanical properties of the film and affect the structural stability of the foam.
An approach that has been developed for preparing a dielectric polymer foam with pore sizes in the nanometer regime involves the use of block copolymers composed of a high temperature, high Tg polymer and a second component that can undergo clean thermal decomposition with the evolution of gaseous by-products to form a closed-cell, porous structure. See, e.g., Hedrick et al. (1993) Polymer 34:4717, and Hedrick et al. (1995) Polymer 36:4855. The process involves the use of block copolymers that can undergo thermodynamically controlled phase separation to provide a matrix with a dispersed phase having a roughly spherical equilibrium morphology, monodisperse in size and discontinuous. By using as a host or matrix material a thermally stable polymer of low dielectric constant and, as the dispersed phase, a labile polymer that undergoes thermolysis at a temperature below the Tg of the vitrified matrix to yield volatile reaction products, one can prepare foams with pores in the nanometer dimensional regime that have no percolation pathway; they are closed structures with nanometer size pores that contain air.
While the method has proved to be useful in the laboratory, the inventors herein have found formation of porous structures to be problematic in several respects. That is, although the concept was demonstrated in principle (see Hedrick et al. (1993); and Hedrick et al. (1995)), application was complicated by synthetic difficulties and by the extremely small thermal processing window. Also, the thermal stability of the foam product was limited to about 350–375° C. for thermoplastic hosts (Hedrick et al. (1996) J. Polym. Sci.; Polym. Chem. 34, 2867). Furthermore, although dielectric constants of 2.2–2.4 were achieved at porosity levels less than about 20% (see Hedrick et al. (1996)), the pore volume could not be further increased without compromising the small domain sizes and/or the non-interconnectivity of the pore structure.
In U.S. Pat. No. 5,953,627 to Carter et al., also assigned to International Business Machines Corporation (Armonk, N.Y.), low molecular weight amines are described as processing aids and toughening agents to improve the mechanical properties of silsesquioxane films. However, when low molecular weight amines are used as toughening agents or as pore-generating agents (or “porogens”), e.g., in the preparation of organosilicate dielectric materials, formulation stability is lacking and unwanted gelation often occurs within a very short period of time (on the order of hours). U.S. Pat. No. 6,107,357 to Hawker et al., also assigned to International Business Machines Corporation, describes the use of polymeric porogens in the preparation of porous dielectric materials, and while the process is effective, the range of nanoporosities that can be obtained is somewhat limited; that is, the void percentage in the porous product is limited to about 35%.
The present invention provides a novel method for preparing highly porous, low dielectric constant materials comprised of porous matrices with a significantly increased range of obtainable void volumes, wherein the matrices can contain non-interconnected, “closed cell” pores or open celled highly porous channels in the form of sharply defined domains. Typically, the pores and/or porous channels are less than about 500 Å in diameter, preferably less than about 200 Å in diameter, more preferably less than about 100 Å. The porous matrices have very low dielectric constants (on the order of 2.0–2.2, or even lower), are thermally stable at temperatures in excess of about 400° C., preferably greater than about 450° C., and are readily processable by current fabrication techniques, including reactive ion etching (RIE), CMP, and high temperature annealing. The spinning solutions maintain a relatively low viscosity (less than about 10.0 cp) over an extended period (up to 30 days or more). The method involves the use of various nitrogen-containing polymers as pore-generating agents. In view of the problems with formulation stability and gelation encountered by using lower molecular weight, basic nitrogenous molecules in dielectric films, i.e., as in U.S. Pat. No. 5,953,627 to Carter et al., it is indeed surprising that nitrogenous polymers would provide excellent formulation stability. The nitrogenous porogenic polymers are useful not only in the preparation of highly porous, low dielectric constant materials, but also in the preparation of substantially defect-free dielectric coatings, and as toughening agents and coating stabilizers.